结合整体和局部特征的步态识别方法  

Gait Recognition Method Combining Global and Local Features

作　　者：张智 常超伟[3] 王雷[1,2] 刘博 ZHANG Zhi;CHANG Chaowei;WANG Lei;LIU Bo(College of Information Science and Technology,Hebei Agricultural University,Baoding 071000,China;Hebei Key Laboratory of Agricultural Big Data,Baoding 071000,China;North Automatic Control Technology Institute,Taiyuan 030006,China)

机构地区：[1]河北农业大学信息科学与技术学院,河北保定071000 [2]河北省农业大数据重点实验室,河北保定071000 [3]北方自动控制技术研究所,太原030006

出　　处：《火力与指挥控制》2023年第4期141-146,共6页Fire Control & Command Control

基　　金：国家自然科学基金(61972132);河北农业大学自主培养人才科研专项基金资助项目(PY201810)。

摘　　要：针对已有步态识别方法注重对步态整体特征建模,却忽略了包含细节信息的局部特征问题,提出了一种整体与局部特征相结合的双分支步态集合特征学习网络。该方法以步态轮廓集合作为输入,通过双分支网络提取两种空间尺度的步态特征,利用特征映射模块将两种特征融合得到步态表示。通过在步态数据集CASIA-B中进行了跨视角识别实验,结果表明该方法在正常行走、背包和穿大衣条件下相比主流方法Rank-1准确率取得了有效提升。Aiming at the existing gait recognition methods that focus on modeling the global features of the gait,but ignore the local features containing detailed information,a two-branch gait set feature learning network combining global and local features is proposed.The method takes a gait silhouette set as input,extracts the gait features of two spatial scales through a two-branch network,and finally uses the feature mapping module to fuse the two features to obtain the gait representation.A series of cross-view recognition experiments are conducted on the CASIA-B gait dataset.The results show that the Rank-1 accuracy of the proposed method is effectively improved compared with the main stream methods under normal walking,carrying bags and wearing overcoats conditions.

关 键 词：步态识别 整体特征 局部特征 跨视角识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]